By Force of Habit and Cyclical Leverage

Abstract

I show that most of Campbell and Cochrane's (1999) main predictions about asset price dynamics result from an inaccuracy in their model's numerical solution. In the accurate solution, volatility is a hump-shaped function of the model's state and falls in recessions. The leverage effect counterfactually disappears in recessions. Risk premia are substantially less cyclical and returns less predictable than reported in the original paper. At the 7-year horizon, the predictive R-squared falls from 30\% for the original solution to 10\% for the accurate solution. However, I show that the original findings can be restored by augmenting the model with countercyclical leverage.

Keywords: External habits, countercyclical volatility, return predictability, leverage, numerical solution methods

JEL Classification: G12, G13, G33